Bandwidth Prediction in TDM-PON-based Mobile Fronthaul for Small Cell CRAN

The increase in demand of high bandwidth has led to the evolution of next generation passive optical network (NG-PON) such as 10-gigabit passive optical network (XG-PON). Not only this, the NG-PON is becoming the popular choice for mobile xhaul for small cell centralized/cloud radio access network (CRAN). These networks run a bandwidth management module at central office to manage the available bandwidth. The efficiency of the module can be increased if the upcoming bandwidth demand is known in advance. In this paper, an existing forecasting model known as long-short term memory (LSTM) model, also known as Vanilla LSTM model is used for predicting the bandwidth demand from an optical network unit (ONU) or remote radio heads (RRH). In addition to this, the results is also obtained for other LSTM models such as stacked LSTM and Bi-directional LSTM. The simulation results are compared in terms of various performance metrics and the analysis confirmed the superiority of vanilla LSTM over other LSTM variants for predicting the bandwidth of an ONU/RRH.